Regression to the Mean and the Cover Jinx: Posterior Calibration of Fortune Reports and Serial Independence in the Tale of the Lost Horse: chapter answers
Owners widely believe that when the best store or creator does worse the next year it is because the award made them complacent, and that when the worst store improves after criticism the criticism worked. This book writes that common sense as runnable arithmetic and tests when it fails. The core model is the regression line of the bivariate normal: given this year's score, next year's conditional expectation is the correlation times this year's score. Its stores, creators, customers and reports are illustrative; all calculations are generated by its programs; the book is general business ana…
DOI pending · monograph page goes live after publication · 5DL-B-202 · 中文版
- Chapter 1 · A Conditional Expectation with Slope Below One: Why Top Stores Slip the Next Year
Chapter 1 shows that with a year-to-year correlation of 0.5 a share of 0.8380 of top-decile stores score lower next year although nobody relaxed, and rebuilds Secrist's "triumph of mediocrity" with Hotelling's rebuttal. - Chapter 2 · The Flight Instructor's Illusion: Why Criticism Looks More Effective than Praise
Chapter 2 replays the flight instructor's season: with feedback that does nothing, the instructor still sees criticism outperform praise by 2.529 units, and only a randomised comparison corrects it. - Chapter 3 · Splitting Variance into Skill and Luck: Repeated Measures, Intraclass Correlation and Reliability
The correlation between two years is the share of the total variance that is skill, and it can be estimated from repeated measurements by a one-way random-effects analysis of variance; but the estimate is not precise, a… - Chapter 4 · Shrinkage Estimation: James and Stein, Efron and Morris's Eighteen Batters, and Shrinkage Ranking
Chapter 4 explains shrinkage with Efron and Morris's eighteen batters: in a simplified version the total squared error of the raw averages is 3.53 times that of the shrunk estimates. - Chapter 5 · The Cover Jinx: Select at the Peak and You Will See a Fall
Chapter 5 shows that the cover jinx is a product of selecting at the peak: a winner chosen from a hundred gets worse in the next period with probability 0.918 even when the cover does nothing, and offers a residual test… - Chapter 6 · Posterior Calibration of Fortune Reports: Buying at the Low Point, Barnum Statements and the Brier Score
For an arbitrary month the share of next months that are better is 0.500, just as symmetry requires. But with $\rho$ at 0.3 and purchases only in the worst 0.05 of months, the next month is better in a share of 0.924, w… - Chapter 7 · Sources East and West: the Huainanzi's Old Man Who Lost His Horse and Galton's 1886 "Regression towards Mediocrity"
Chapter 7 reads the Huainanzi's tale of the lost horse alongside Galton's 1886 paper and computes the bias of the "blessing after blessing" share in short records, 0.4048 for four entries. - Chapter 8 · Closing the Loop: Annual Reviews by Shrinkage Ranking, Copy without Causal Overstatement, and a Pre-registered Plan
Chapter 8 gives an annual review by shrinkage ranking (the first-placed store in the example has only 0.634 probability of a top-three finish), a copy audit, a pre-registered plan and the distilled Shrinkage Ranking Cal…
Illustrative model calculation, not a live quote